Outlier removal excludes observations that sit unusually far from the rest of a dataset. It can make a pattern look cleaner, but removing unusual observations too aggressively can distort the claim. section 1.4
Treat an outlier removal as a question, not an automatic cleanup step: what makes the observation unusual, and what conclusion changes when it disappears? section 1.4
When reviewing a statistical claim:
- Check whether unusual observations were removed too zealously. section 1.4
- Consider the removal alongside sampling bias, process assumptions, and ignored variables. section 1.4
- Separate a statistically significant result from a trustworthy explanation of what caused it. section 1.4
